SWE-bench Multilingual
Multilingual SWE-bench
codingactive
Score in %
Definition
- Version
- not in the notes
- Category
- coding
- Maintainer
- SWE-bench team, USA
- Status
- active
- Home page
- https://www.swebench.com/
Sources
- Introducing Composer 2 , Cursor cursor.com https://cursor.com/blog/composer-2
- huggingface.co/mistralai/Devstral-2-123B-Instruct-2512 https://huggingface.co/mistralai/Devstral-2-123B-Instruct-2512
- SWE-bench Leaderboards swebench.com https://www.swebench.com/
Scores at a Glance
Score in %
Higher is better for this unit.
1
Composer 2Anysphere (Cursor)
Devstral 2 (123B, 2512)Mistral AIDevstral Small 2 (24B, 2512)Mistral AI
Composer 2Anysphere (Cursor)73.7 %
vendor-reported2Devstral 2 (123B, 2512)Mistral AI61.3 %
vendor-reported3Devstral Small 2 (24B, 2512)Mistral AI55.7 %
vendor-reported0255075100
Where this comes from
Each bar is one model, with its most independent recorded score. Hatched bars are vendor-reported; the word beside each bar says who ran it. Settings and dates differ between rows; the Scores table below has every result.
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
| Composer 2 | 73.7 % | vendor-reported | – | – | – | cursor.com |
 Ling 3.0 Flash (Ling-3.0-flash) | 72.4 % | vendor-reported | – | Ling-3.0-flash model card eval results | Jul 2026 | huggingface.co |
 MiMo-V2-Flash | 71.7 % | vendor-reported | – | MiMo-V2-Flash model card, post-trained model table | 16 Dec 2025 | huggingface.co |
| Devstral 2 (123B, 2512) | 61.3 % | vendor-reported | – | – | – | huggingface.co |
 DeepSeek-V3.1-Terminus | 57.8 % | vendor-reported | Model card comparison table; mode not stated | DeepSeek-V3.1-Terminus model card | 22 Sep 2025 | huggingface.co |
| Devstral Small 2 (24B, 2512) | 55.7 % | vendor-reported | – | – | – | huggingface.co |
| DeepSeek-V3.1 | 54.5 % | vendor-reported | Non-thinking, agent mode | DeepSeek-V3.1 model card | 21 Aug 2025 | huggingface.co |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "SWE-Bench Multilingual", https://aurion.semai.id/benchmarks/swe-bench-multilingual/, records of October 2026.